CloudCrowd can share code with your Rails applications, taking advantage of pre-existing models, and performing expensive database operations in parallel. The precise technique to accomplish this may vary — not all applications will need to load the entire Rails stack in order to perform a job. At DocumentCloud, we store our actions in app/actions, and load the Rails environment in order to seamlessly access all of our models, just like during a real web request. Because CloudCrowd forks a worker for each work unit, Rails is only loaded once when a node spins up, and the child processes don’t incur the expense of loading it.
Here’s an example snippet of code that we use to setup the Rails environment for an action. The CloudCrowd.node? bit ensures that the Rails stack doesn’t ever get loaded on the central server.
# Inherit Rails environment from Sinatra. RAILS_ROOT = File.expand_path(Dir.pwd) RAILS_ENV = ENV['RAILS_ENV'] = ENV['RACK_ENV'] # Load the DocumentCloud environment if we're in a Node context. if CloudCrowd.node? require 'rubygems' require 'activerecord' ActiveRecord::Base.logger = Logger.new(STDOUT) require 'config/environment' end